We can now manipulate matter at the level of individual atoms and molecules, and we are beginning to see some of the results of this nanotechnology in the form of useful products. But the most sophisticated results of nanotechnology - working machines, devices and systems made on the molecular scale - have yet to be constructed. If such machines can be made, they will undoubtedly make possible great advances in medicine, energy and information technology, but what kind of engineering principles will they be based on? Our natural tendency is to assume that nanoscale machines will operate on the same principles as human-scale engineering, but physics looks different on the nanoscale in ways that will make this approach very difficult. The most sophisticated nanoscale machines and devices we know about now are the sub-cellular machines, made of natural polymers such as proteins and nucleic acids, which make all life work. This natural nanotechnology is based on quite different design principles to the principles we learn in macroscopic engineering. The components of the machines are soft and floppy, and the way they works relies on features of the physics of the nanoscale - like the constant shaking of Brownian motion and the universal stickiness that arises from strong surface forces - which have no analogue at the macroscale. It follows that, in designing synthetic nanoscale machines, we should learn from the way nature exploits the special physics of the nanoscale, rather than trying to design around the different features of nanoscale physics, as we would end up doing if we tried to apply macroscopic engineering design principles. In brief, nanotechnology should look more like biology than engineering.